877884 - Table creation probably generates wrong primary key

Status: VERIFIED FIXED

(Cloud Services Graveyard :: Server: Sync, defect, P2)

Description

[Toby Elliott [:telliott]]

http://hg.mozilla.org/services/server-storage/file/cffa43a1ceea/syncstorage/storage/sqlmappers.py#l77

The table definition defines id, username and collection columns and designates them PK. However, the PK should be username, collection, id (since username, collection is a common query).

Comment 1

[Ryan Kelly [:rfkelly]]

This is not a problem in current prod because ops create the tables by hand.  But let's fix it in both places in case that situation changes in future.

Comment 2

[Ryan Kelly [:rfkelly]]

Attachment: sync1.1-wbo-primary-key-order.diff

Comment 3

[Ryan Kelly [:rfkelly]]

Attachment: sync1.5-bso-primary-key-order.diff

Comment 4

[Toby Elliott [:telliott]]

Comment on attachment 8388958 [details] [diff] [review]
sync1.5-bso-primary-key-order.diff

Review of attachment 8388958 [details] [diff] [review]:
-----------------------------------------------------------------

This, btw, is another reason I'm not a fan of ORMs :P

Comment 5

[Ryan Kelly [:rfkelly]]

https://github.com/mozilla-services/server-syncstorage/commit/6c004bc35ec951b2aadf31e9ea3e93b88705e34c
http://hg.mozilla.org/services/server-storage/rev/617af5cd31e3

Comment 6

[James Bonacci [:jbonacci]]

:rfkelly - I verified this in code, but what do we need to do to verify the following:
"But let's fix it in both places in case that situation changes in future."

Comment 7

[Ryan Kelly [:rfkelly]]

> "But let's fix it in both places in case that situation changes in future."

Just make sure it's in both sync1.5 and sync1.1 codebases

Comment 8

[James Bonacci [:jbonacci]]

Done.